Re: [saag] Revision of "Attacks on Cryptographic Hashes in Internet Protocols"

[Paul Hoffman <[e-mail]>]

On Nov 14, 2012, at 5:14 AM, David McGrew (mcgrew) <[e-mail]> wrote:

> I think you (Joe) have a valid point about how the draft could be
> improved, though the draft does address the use question somewhat (it even
> has a section on "How Internet Protocols Use Hash Algorithms").

Yes, it does.

> I have a suggestion: the draft could
> 
> 1) more precisely define the different ways that hash functions are used
> in more detail (in signatures, in HMAC, KDFs, other message authentication
> codes, integrity checking, ...)   The definitions should be clear enough
> that a relative crypto novice, looking at a specification that describes a
> use of a hash function, could correctly categorize that use.

Proposed wording would be greatly appreciated here. I cannot see how to add that text and have it be anything other than singing to the choir.

> 2) relate the security of each use case to the collision/first
> preimage/second preimage attacks

Ditto here. When we tried this seven years ago, we were attacked for being to restrictive in our descriptions. Seriously: if you have a contribution to make that you think is readable to a relative crypto novice and still accurate, we're all ears.

> 3) have a section that describes uses of hash functions in Internet
> protocols that rely on collision resistance.   (My thinking here is that
> there are many uses of hash functions, and so we should focus on the most
> security critical cases)

We thought we had that in the existing RFC and the current draft. Which other protocols are you thinking of?

--Paul Hoffman